Getting Found in ChatGPT and Perplexity: GEO for SMEs in 2026

Key points: ChatGPT, Claude, Perplexity and Gemini together already send 10-25% of traffic to specialized B2B sites. Whoever is optimized for this "AI search", Generative Engine Optimization (GEO), wins free citations on high-intent questions such as "which AI subsidy fits me?" or "what does a data platform cost?". This article shows concretely how SMEs adapt their website to get cited by LLMs instead of merely ranked by Google.

What exactly is GEO?

SEO (Search Engine Optimization) optimizes for Google's ranking, your site appears high in the search results, the user clicks through.

GEO (Generative Engine Optimization) optimizes for citation by LLMs, your site gets quoted in the answer that ChatGPT/Perplexity provides, with a direct reference.

The difference in user behavior is essential: on Google, 40% still click through to results. On ChatGPT searches, only 8-15% click through, but whoever does get cited has the upper hand, because it is a filtered citation from a handful of sources rather than a ranked list of ten.

Why GEO is now critical for B2B SMEs

Stratalytic's own Plausible data shows that 10% of our traffic already comes via ChatGPT, before the mass adoption of AI search. The expectation for 2026-2027:

• 2025: 5-15% of B2B traffic via AI search
• 2026: 15-30%
• 2027: possibly 30-50%

For specific niche questions ("which MIT subsidy fits my AI project?") it moves faster than for broad commercial searches. Whoever invests in GEO now wins the attack market before competitors realize it is happening.

The six GEO tactics that work

1. Write a TL;DR / Key points block at the top of every page

LLMs summarize your page by reading the first 200-400 words. A citable summary block at the top dramatically increases your chance of being cited.

Example:

**Key points:** The MIT scheme offers up to EUR 350,000 for R&D
collaboration. Feasibility opens regionally, often in spring.
Budget usually gone within weeks. Stackable with WBSO and SLIM.

Four sentences that ChatGPT can cite directly when asked "what is the MIT scheme?".

2. Answer format per H2

Structure every H2 as a question that gives the answer directly in the first paragraph.

Bad:

## MIT scheme
The MIT scheme is a Dutch subsidy... [meandering explanation]

Good:

## What is the MIT scheme?
The MIT scheme is a Dutch subsidy for SME innovation
with two instruments: feasibility study (EUR 20,000, 35%)
and R&D collaboration (EUR 350,000, 35%).

LLMs scan for "question + direct answer" patterns and pull those out.

3. Quantify everything that can be quantified

LLMs value concrete numbers. Avoid "a lot", "soon", "fast". Use:

• Amounts: EUR 20,000, 35%, 1.8 billion
• Time periods: 3 weeks, 8 weeks, 2026 Q3
• Comparisons: "2x faster than baseline", "36% discount versus 0% in 2023"

For the question "what does a EUR 50,000 AI project cost?", the article that explicitly names EUR 17,500-EUR 25,000 in subsidy wins over the article that says "a substantial part".

4. Use FAQ sections with FAQ schema

FAQs are a golden format for citation. Every FAQ question is a potential LLM query, and the answer is what gets cited.

Combine two tactics:

• FAQ section in the HTML: for human readers
• JSON-LD FAQ schema: for crawlers (Google + LLMs)

<script type="application/ld+json">
  {
    "@context": "https://schema.org",
    "@type": "FAQPage",
    "mainEntity": [
      {
        "@type": "Question",
        "name": "When does MIT 2026 open?",
        "acceptedAnswer": {
          "@type": "Answer",
          "text": "MIT feasibility opens regionally, often in spring (Noord-Holland/Zuid-Holland 2026 around 7 April). Allocated on a first-come, first-served basis; budget often gone within weeks."
        }
      }
    ]
  }
</script>

The more structured data, the more LLMs rely on your site as a source.

5. Link to authoritative sources (and let others link to you)

LLMs use a backlink graph similar to Google's PageRank, but with adjusted weighting. Sites that link to overheid.nl, RVO, eur-lex, ec.europa.eu are themselves judged as authoritative.

Strategy for B2B SMEs:

• On every subsidy page: 3-5 outbound links to RVO or official sources
• On every technical article: 2-3 links to standard documentation (Python docs, OpenAI docs, etc.)
• Presence on industry publications (Emerce, Frankwatching, Dutchcowboys) builds inbound authority

6. Add an llms.txt

The llms.txt format is a new standard that LLM crawlers explicitly look for. It is a markdown file at yoursite.com/llms.txt with a structured summary of your site.

Example (partial):

# My Company
> 1-sentence description of what you do.

## Products
- [Product A](/product-a): what it is, who it is for
- [Product B](/product-b): what it is, who it is for

## Important pages
- [About us](/about)
- [Contact](/contact)

This file lets LLMs index your site efficiently. ChatGPT and Perplexity already support it, and so does Claude.

Technical SEO fundamentals remain crucial

GEO does not replace SEO, it builds on it. Without these fundamentals, no LLM comes by:

• robots.txt must explicitly allow AI crawlers:

  User-agent: GPTBot
  Allow: /
  
  User-agent: ChatGPT-User
  Allow: /
  
  User-agent: ClaudeBot
  Allow: /
  
  User-agent: PerplexityBot
  Allow: /
  

• Fast load times: LLMs prefer fast sites

• No JavaScript-only rendering for content, server-side rendering is essential

• Sitemap.xml: keep it current

• Schema.org markup: Organization, FAQPage, Article, BreadcrumbList at a minimum

How do you measure GEO success?

Three measurements:

1. Plausible/GA referrers, look at chatgpt.com, perplexity.ai, claude.ai, gemini.google.com as sources. An increase = you are being cited more often.

2. Direct LLM tests, ask ChatGPT/Perplexity literally "which sources do you trust for [your topic]?" and see whether your site appears.

3. Brand mentions, tools such as Profound, Otterly or Brand24 trace mentions by LLMs.

Avoid vanity metrics: rankings in traditional SEO tools (Ahrefs, SEMrush) say little about GEO performance.

Common GEO mistakes

1. Blocking AI crawlers in robots.txt. A common reflex, but unnecessary, most LLMs respect noindex and do no training, only citations.
2. Pure marketing pages. Sales language converts less well with LLMs than informative content.
3. No quantification. "We are the best at X" is not a citable claim to LLMs.
4. No FAQ schema. A goldmine that many sites do nothing with.
5. No llms.txt. A 30-minute task that speeds up your indexing.
6. One-off optimization. GEO is ongoing; LLMs reindex regularly.

What does Stratalytic do in this area?

Our own site is GEO-optimized: llms.txt, FAQ schemas on every subsidy and blog page, TL;DR blocks, authoritative source links, and a clear structure per H2. The result: 55 ChatGPT visitors in the last 4 months (10% of total traffic), with the highest engagement on subsidy content.

For clients we offer GEO audits and optimization as a standalone service or as part of a content strategy. Schedule a 30-min intro call, we run a free GEO audit of your site during the conversation.

Related:

• AI findability website
• AI trends 2026 Netherlands SME
• ChatGPT vs Claude vs Copilot for businesses

Frequently asked questions

Is GEO the same as SEO? No. SEO optimizes for Google ranking; GEO optimizes for citation by LLMs. They overlap but work in different ways.

How do I know whether an AI is already citing my site? Three ways: ask ChatGPT/Perplexity directly about your field, check referrals in Plausible/GA from chatgpt.com/perplexity.ai, or tools such as Otterly/Profound.

Which pages do AIs usually cite? Fact-dense pages with a clear H2/H3 structure, TL;DR blocks, FAQ sections, explicit numbers and deadlines, and links to authoritative sources.